Advanced Practices in Travel Forecasting (2010)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

5PURPOSE OF STUDY For half a century, transportation modeling has been an estab- lished field of practice for infrastructure planning and policy analysis. The four-step model, which separates trip generation, distribution, mode choice, and assignment, is an established approach widely used in today’s transportation modeling. In the last decade, however, policymakers have begun asking more complex questions, such as about the impacts of road pricing distinguished by vehicle types, occupancy, time-of-day, or level of congestion. How far may rising fuel prices affect travel behavior with regard to mode choice, trip chaining, or the choice of locations for living, working, shopping, and leisure activities? What is the impact of alternative growth scenarios, such as transit-oriented development or smart growth strate- gies, on traffic volumes? How do rising freight volumes impact traffic flows at different time-of-day periods? These and other questions asked by policymakers are difficult if not impossible to answer with traditional modeling approaches; therefore, a new type of modeling was needed. At the same time, science has made significant progress by learning from large-scale projects such as TRANSIMS and other more disaggregated approaches. Since the beginning of this century, more and more agencies have explored the benefits of advanced modeling. It is important to define at the outset what is meant by the term, “advanced modeling.” It undoubtedly means different things to different people. In preparing this Synthesis the definition cited in TRB Special Report 288: Metropolitan Travel Forecasting: Current Practice and Future Direction was followed. That is, advanced modeling generally encom- passes practices beyond the four-step sequential modeling par- adigm traditionally used in travel demand forecasting and its close variants. Although often considered synonymous with tour- and activity-based travel modeling, advanced modeling includes other nontraditional approaches, such as region-wide dynamic traffic assignment and traffic microsimulation, land use transport modeling, freight and commercial vehicle mod- eling, and explicit linkages to statewide travel models. Each is discussed in this report. A few other equally compelling practices, such as mobile source emissions modeling, carbon footprinting, and analysis of greenhouse gas effects arguably fall into the same category, but were not examined as part of this work. Being a relatively new field of practice, two questions about advanced modeling are central to this report. The first question explores the reasons why agencies may want to move to advanced models. If a traditional model is estab- lished and major efforts both in budget and time are required to move to advanced models, it is important that an agency explore carefully the rationale for abandoning a working traditional model and moving to advanced modeling tech- niques. This report explores the limits of traditional models and highlights policy questions that may motivate switching to advanced models. The findings may help agencies dis- cover the most suitable approach for every modeling task. Second, common obstacles encountered when moving toward advanced models are listed and analyzed. By addressing the obstacles many advanced modeling projects have encoun- tered, those agencies that decide to move toward advanced models may be able to circumvent some difficulties faced by other agencies. Topics addressed included institutional issues, funding, project organization, and the technical implementa- tion. The literature reviewed supports some findings with a theoretical background. Ultimately, stimulating the discussion about advanced models in general is an underlying purpose of this report. Even if some agencies—for various reasons—decide to continue using traditional models, an intensified discussion about the advantages and shortcomings of different modeling approaches can do nothing but improve everyone’s compe- tence in transportation modeling. STUDY METHODOLOGY A comprehensive review of current and past efforts in advanced modeling informed this report. A list of all known existing modeling projects in North America was assem- bled at the outset of the project, based on study team knowl- edge of or participation in the work, review of the literature and recent conference proceedings, and leads from TRB staff and the topic panel. These known projects were cate- gorized by their progress to date, as shown in Table 1. Port- land (Oregon) is somewhat unique in this regard, as it has the experience of having abandoned earlier efforts, only to start anew several years later. The Oregon Department of Transportation (DOT) has progressed to a second genera- tion of models, but the rest of the advanced models sum- marized in the table are first generation efforts. Under- standing these projects and the lessons learned were the focus of this study. CHAPTER ONE INTRODUCTION

6TABLE 1 AGENCIES SELECTED FOR FURTHER STUDY (AS OF AUGUST 2009) DTA = dynamic traffic assignment. The study team interviewed more than 30 practitioners and researchers during the course of this study. Most were selected because of their involvement in the implementa- tion, application, or evaluation of one or more of the efforts listed in Table 1. Others were chosen because of their work in research and development in advanced modeling or rec- ognized expertise in one of the subareas identified earlier. It was hoped that these individuals would be able to share success stories and important lessons learned from developing and using such models within metropolitan planning organizations (MPOs). An interview approach was chosen instead of a ques- tionnaire because of the flexibility afforded, as those contacted covered a wide spectrum of experiences. Many of the issues were complex and better understood through discussion. The interviews were conducted by phone or in person, with the latter being carried out when there were others reasons for study team members to be in the area. Participants were first contacted by e-mail, which was accompanied by an Interview Preparation Guide consisting of questions in five broad topic areas. The Guide is shown in Appendix A. Although the inter- viewers used the Guide to orient the discussion, the respon- dents were not asked to return the completed Guide, rather they were asked to make notes on it as they thought about the questions. It was hoped that over the several weeks between being initially contacted and interviewed that the respondents would have several occasions to add thoughts or notes that would help with recall during the interview process. The interviews typically lasted between one and three hours each. The interviewers identified and initially focused on the topics the respondent was most interested in or had the most to say about. Respondents were asked to rank a list of issues in the Guide. These rankings helped orient the interview toward those topics the respondent believed were either not significant or important to them. Toward the end of the inter- view the interviewer typically asked for responses to those questions not already covered. In some instances a second or third contact was required to obtain information about the cost of model development and implementation, information eagerly sought by many practitioners and agencies contem- plating a move to advanced modeling. The study team discussed the highlights of each interview, with the interviewer(s) summarizing the key topics and discus- sion items. As the number of completed interviews increased, most of the effort was devoted to identifying recurring and significant issues and themes from the collective responses. Those major findings defined the structure and content of the following chapters. ORGANIZATION OF REPORT The findings of this synthesis are summarized in the next five chapters. Following this introductory chapter, chapter two provides a description of current advanced modeling practices

7across the United States. The efforts described were summa- rized from the literature, unpublished documentation, online searches, and the interviews carried out as part of this proj- ect. The focus is on emerging and operational models in use by MPOs and as such does not attempt to encompass the large amount of parallel academic work. Chapter three describes the benefits ascribed to advanced models by their users. Their advantages over other analytical approaches, to include traditional modeling practices, are discussed at length. Although the benefits of advanced mod- els are taken for granted by their proponents, considerable debate about the case for moving to them persists among prac- titioners. Many practitioners noted that their criteria for acceptance of such models were different than for academics. This chapter attempts to build the case for advanced models in the words of the practitioners. Implementation and institutional issues faced by practi- tioners are highlighted in chapter four. These issues were gleaned from the review of practice and the interviews, and include a variety of methodological, data, cost, and institu- tional issues unique to practitioners. This chapter describes some of the barriers that have been encountered and how the various agencies dealt with them. Issues unique to a particular agency are documented, because they might apply to others reading this report. However, the primary focus has been to deal with recurrent themes across most of the agencies. Suc- cess stories are highlighted where available. Chapter five presents the lessons learned by the developers and users of these advanced models. During the interviews this was couched as the following open-ended question: “If you had it to do over, what would you have done differently?” Most respondents reported that they would have followed substantially the same path, but all were able to point to a few or more ideas or improvements. Many were already adapting their work to take advantage of their new insights. The degree to which these lessons are broadly applicable or widely trans- ferable remains to be seen. It is interesting to note that most respondents asked during the interview what lessons had already been reported and largely agreed with the wisdom shared by their contemporaries. Chapter six presents several case studies that underscore and expand on some of the primary themes presented in earlier chapters. Each illustrates a slightly different view on advanced modeling: • San Francisco was one of the first activity-based models put into practice, and has the largest number of applica- tions in real-world studies. • Sacramento recently completed the implementation of an activity-based travel model that has been used for studies of the effect of the built environment on travel. • Portland (Oregon) was also an early leader in activity- based modeling that switched its focus to TRANSIMS and is now approaching activity-based modeling again. • Lake Tahoe’s activity-based travel model was imported from Columbus (Ohio), making it an interesting study in model transferability. Chapter seven summarizes the major findings from the study and recounts the next steps suggested or desired by the respondents. As such, it offers more of the collective ideas on how to move advanced modeling forward than it does the views of the study team. The references and a list of abbreviations are provided at the end of the report. The Interview Preparation Guide is included as Appendix A and the list of survey respondents is shown in Appendix B.